DonorsChoose¶
DonorsChoose.org receives hundreds of thousands of project proposals each year for classroom projects in need of funding. Right now, a large number of volunteers is needed to manually screen each submission before it's approved to be posted on the DonorsChoose.org website.
Next year, DonorsChoose.org expects to receive close to 500,000 project proposals. As a result, there are three main problems they need to solve:
- How to scale current manual processes and resources to screen 500,000 projects so that they can be posted as quickly and as efficiently as possible
- How to increase the consistency of project vetting across different volunteers to improve the experience for teachers
- How to focus volunteer time on the applications that need the most assistance
The goal of the competition is to predict whether or not a DonorsChoose.org project proposal submitted by a teacher will be approved, using the text of project descriptions as well as additional metadata about the project, teacher, and school. DonorsChoose.org can then use this information to identify projects most likely to need further review before approval.
About the DonorsChoose Data Set¶
The train.csv data set provided by DonorsChoose contains the following features:
| Feature | Description |
|---|---|
project_id |
A unique identifier for the proposed project. Example: p036502 |
project_title |
Title of the project. Examples:
|
project_grade_category |
Grade level of students for which the project is targeted. One of the following enumerated values:
|
project_subject_categories |
One or more (comma-separated) subject categories for the project from the following enumerated list of values:
Examples:
|
school_state |
State where school is located (Two-letter U.S. postal code). Example: WY |
project_subject_subcategories |
One or more (comma-separated) subject subcategories for the project. Examples:
|
project_resource_summary |
An explanation of the resources needed for the project. Example:
|
project_essay_1 |
First application essay* |
project_essay_2 |
Second application essay* |
project_essay_3 |
Third application essay* |
project_essay_4 |
Fourth application essay* |
project_submitted_datetime |
Datetime when project application was submitted. Example: 2016-04-28 12:43:56.245 |
teacher_id |
A unique identifier for the teacher of the proposed project. Example: bdf8baa8fedef6bfeec7ae4ff1c15c56 |
teacher_prefix |
Teacher's title. One of the following enumerated values:
|
teacher_number_of_previously_posted_projects |
Number of project applications previously submitted by the same teacher. Example: 2 |
* See the section Notes on the Essay Data for more details about these features.
Additionally, the resources.csv data set provides more data about the resources required for each project. Each line in this file represents a resource required by a project:
| Feature | Description |
|---|---|
id |
A project_id value from the train.csv file. Example: p036502 |
description |
Desciption of the resource. Example: Tenor Saxophone Reeds, Box of 25 |
quantity |
Quantity of the resource required. Example: 3 |
price |
Price of the resource required. Example: 9.95 |
Note: Many projects require multiple resources. The id value corresponds to a project_id in train.csv, so you use it as a key to retrieve all resources needed for a project:
The data set contains the following label (the value you will attempt to predict):
| Label | Description |
|---|---|
project_is_approved |
A binary flag indicating whether DonorsChoose approved the project. A value of 0 indicates the project was not approved, and a value of 1 indicates the project was approved. |
Notes on the Essay Data¶
-
Prior to May 17, 2016, the prompts for the essays were as follows:
- __project_essay_1:__ "Introduce us to your classroom"
- __project_essay_2:__ "Tell us more about your students"
- __project_essay_3:__ "Describe how your students will use the materials you're requesting"
- __project_essay_3:__ "Close by sharing why your project will make a difference"
-
Starting on May 17, 2016, the number of essays was reduced from 4 to 2, and the prompts for the first 2 essays were changed to the following:
- __project_essay_1:__ "Describe your students: What makes your students special? Specific details about their background, your neighborhood, and your school are all helpful."
- __project_essay_2:__ "About your project: How will these materials make a difference in your students' learning and improve their school lives?"
For all projects with project_submitted_datetime of 2016-05-17 and later, the values of project_essay_3 and project_essay_4 will be NaN.
In [1]:
%matplotlib inline
import warnings
warnings.filterwarnings("ignore")
import sqlite3
import pandas as pd
import numpy as np
import nltk
import string
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.feature_extraction.text import TfidfTransformer
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.metrics import confusion_matrix
from sklearn import metrics
from sklearn.metrics import roc_curve, auc
from nltk.stem.porter import PorterStemmer
import re
import string
from nltk.corpus import stopwords
from nltk.stem import PorterStemmer
from nltk.stem.wordnet import WordNetLemmatizer
from gensim.models import Word2Vec
from gensim.models import KeyedVectors
import pickle
from tqdm import tqdm
import os
from plotly import plotly
import plotly.offline as offline
import plotly.graph_objs as go
offline.init_notebook_mode()
from collections import Counter
1.1 Reading Data¶
In [105]:
project_data = pd.read_csv('train_data.csv',nrows = 60000)
resource_data = pd.read_csv('resources.csv')
In [106]:
print("Number of data points in train data", project_data.shape)
print('-'*50)
print("The attributes of data :", project_data.columns.values)
In [107]:
print("Number of data points in train data", resource_data.shape)
print(resource_data.columns.values)
resource_data.head(2)
Out[107]:
1.2 preprocessing of project_subject_categories¶
In [108]:
catogories = list(project_data['project_subject_categories'].values)
# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039
# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/
# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string
# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python
cat_list = []
for i in catogories:
temp = ""
# consider we have text like this "Math & Science, Warmth, Care & Hunger"
for j in i.split(','): # it will split it in three parts ["Math & Science", "Warmth", "Care & Hunger"]
if 'The' in j.split(): # this will split each of the catogory based on space "Math & Science"=> "Math","&", "Science"
j=j.replace('The','') # if we have the words "The" we are going to replace it with ''(i.e removing 'The')
j = j.replace(' ','') # we are placeing all the ' '(space) with ''(empty) ex:"Math & Science"=>"Math&Science"
temp+=j.strip()+" " #" abc ".strip() will return "abc", remove the trailing spaces
temp = temp.replace('&','_') # we are replacing the & value into
cat_list.append(temp.strip())
project_data['clean_categories'] = cat_list
project_data.drop(['project_subject_categories'], axis=1, inplace=True)
from collections import Counter
my_counter = Counter()
for word in project_data['clean_categories'].values:
my_counter.update(word.split())
cat_dict = dict(my_counter)
sorted_cat_dict = dict(sorted(cat_dict.items(), key=lambda kv: kv[1]))
1.3 preprocessing of project_subject_subcategories¶
In [109]:
sub_catogories = list(project_data['project_subject_subcategories'].values)
# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039
# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/
# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string
# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python
sub_cat_list = []
for i in sub_catogories:
temp = ""
# consider we have text like this "Math & Science, Warmth, Care & Hunger"
for j in i.split(','): # it will split it in three parts ["Math & Science", "Warmth", "Care & Hunger"]
if 'The' in j.split(): # this will split each of the catogory based on space "Math & Science"=> "Math","&", "Science"
j=j.replace('The','') # if we have the words "The" we are going to replace it with ''(i.e removing 'The')
j = j.replace(' ','') # we are placeing all the ' '(space) with ''(empty) ex:"Math & Science"=>"Math&Science"
temp +=j.strip()+" "#" abc ".strip() will return "abc", remove the trailing spaces
temp = temp.replace('&','_')
sub_cat_list.append(temp.strip())
project_data['clean_subcategories'] = sub_cat_list
project_data.drop(['project_subject_subcategories'], axis=1, inplace=True)
# count of all the words in corpus python: https://stackoverflow.com/a/22898595/4084039
my_counter = Counter()
for word in project_data['clean_subcategories'].values:
my_counter.update(word.split())
sub_cat_dict = dict(my_counter)
sorted_sub_cat_dict = dict(sorted(sub_cat_dict.items(), key=lambda kv: kv[1]))
In [110]:
# how to replace elements in list python: https://stackoverflow.com/a/2582163/4084039
cols = ['Date' if x=='project_submitted_datetime' else x for x in list(project_data.columns)]
#sort dataframe based on time pandas python: https://stackoverflow.com/a/49702492/4084039
project_data['Date'] = pd.to_datetime(project_data['project_submitted_datetime'])
project_data.drop('project_submitted_datetime', axis=1, inplace=True)
project_data.sort_values(by=['Date'], inplace=True)
# how to reorder columns pandas python: https://stackoverflow.com/a/13148611/4084039
project_data = project_data[cols]
project_data.head(2)
Out[110]:
In [111]:
#preprocessing project_grade_categories
#project_grade contain grade with space so we remove the grades from the data set
project_grade= list(project_data['project_grade_category'].values)
# remove special characters from list of strings python:
# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/
grade_cat_list = []
for i in project_grade:
# consider we have text like this:
for j in i.split(' '): # # split by space
j=j.replace('Grades','')# clean grades from the row
grade_cat_list.append(j.strip())
project_data['clean_grade'] = grade_cat_list
project_data.drop(['project_grade_category'], axis=1, inplace=True)
my_counter = Counter()
for word in project_data['clean_grade'].values:
my_counter.update(word.split())
project_grade_category_dict= dict(my_counter)
sorted_project_grade_category_dict = dict(sorted(project_grade_category_dict.items(), key=lambda kv: kv[1]))
In [112]:
project_data.head(2)
Out[112]:
1.3 Text preprocessing¶
In [113]:
# merge two column text dataframe:
project_data["essay"] = project_data["project_essay_1"].map(str) +\
project_data["project_essay_2"].map(str) + \
project_data["project_essay_3"].map(str) + \
project_data["project_essay_4"].map(str)
In [114]:
# printing some random reviews
print(project_data['essay'].values[0])
print("="*50)
print(project_data['essay'].values[150])
print("="*50)
print(project_data['essay'].values[1000])
print("="*50)
print(project_data['essay'].values[20000])
In [115]:
# https://stackoverflow.com/a/47091490/4084039
import re
def decontracted(phrase):
# specific
phrase = re.sub(r"won't", "will not", phrase)
phrase = re.sub(r"can\'t", "can not", phrase)
# general
phrase = re.sub(r"n\'t", " not", phrase)
phrase = re.sub(r"\'re", " are", phrase)
phrase = re.sub(r"\'s", " is", phrase)
phrase = re.sub(r"\'d", " would", phrase)
phrase = re.sub(r"\'ll", " will", phrase)
phrase = re.sub(r"\'t", " not", phrase)
phrase = re.sub(r"\'ve", " have", phrase)
phrase = re.sub(r"\'m", " am", phrase)
return phrase
In [119]:
# https://gist.github.com/sebleier/554280
# we are removing the words from the stop words list: 'no', 'nor', 'not'
stopwords= ['i', 'me', 'my', 'myself', 'we', 'our', 'ours', 'ourselves', 'you', "you're", "you've",\
"you'll", "you'd", 'your', 'yours', 'yourself', 'yourselves', 'he', 'him', 'his', 'himself', \
'she', "she's", 'her', 'hers', 'herself', 'it', "it's", 'its', 'itself', 'they', 'them', 'their',\
'theirs', 'themselves', 'what', 'which', 'who', 'whom', 'this', 'that', "that'll", 'these', 'those', \
'am', 'is', 'are', 'was', 'were', 'be', 'been', 'being', 'have', 'has', 'had', 'having', 'do', 'does', \
'did', 'doing', 'a', 'an', 'the', 'and', 'but', 'if', 'or', 'because', 'as', 'until', 'while', 'of', \
'at', 'by', 'for', 'with', 'about', 'against', 'between', 'into', 'through', 'during', 'before', 'after',\
'above', 'below', 'to', 'from', 'up', 'down', 'in', 'out', 'on', 'off', 'over', 'under', 'again', 'further',\
'then', 'once', 'here', 'there', 'when', 'where', 'why', 'how', 'all', 'any', 'both', 'each', 'few', 'more',\
'most', 'other', 'some', 'such', 'only', 'own', 'same', 'so', 'than', 'too', 'very', \
's', 't', 'can', 'will', 'just', 'don', "don't", 'should', "should've", 'now', 'd', 'll', 'm', 'o', 're', \
've', 'y', 'ain', 'aren', "aren't", 'couldn', "couldn't", 'didn', "didn't", 'doesn', "doesn't", 'hadn',\
"hadn't", 'hasn', "hasn't", 'haven', "haven't", 'isn', "isn't", 'ma', 'mightn', "mightn't", 'mustn',\
"mustn't", 'needn', "needn't", 'shan', "shan't", 'shouldn', "shouldn't", 'wasn', "wasn't", 'weren', "weren't", \
'won', "won't", 'wouldn', "wouldn't"]
1.4 Preprocessing of `project_title`
In [122]:
print(project_data['project_title'].values[0])
print("="*50)
print(project_data['project_title'].values[150])
print("="*50)
print(project_data['project_title'].values[1000])
print("="*50)
print(project_data['project_title'].values[20000])
print("="*50)
In [125]:
# we cannot remove rows where teacher prefix is not available therefore we are replacing 'nan' value with
# 'null'(string)
#https://stackoverflow.com/questions/42224700/attributeerror-float-object-has-no-attribute-split
project_data['teacher_prefix'] = project_data['teacher_prefix'].fillna('null')
1.5 Preparing data for models¶
In [103]:
project_data.columns
Out[103]:
we are going to consider
- school_state : categorical data
- clean_categories : categorical data
- clean_subcategories : categorical data
- project_grade_category : categorical data
- teacher_prefix : categorical data
- project_title : text data
- text : text data
- project_resource_summary: text data (optinal)
- quantity : numerical (optinal)
- teacher_number_of_previously_posted_projects : numerical
- price : numericalSplitting data into Train and cross validation(or test): Stratified Sampling¶
In [129]:
# train test split
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(project_data, project_data['project_is_approved'],stratify = project_data['project_is_approved'], test_size=0.33)
X_train,X_cv, y_train, y_cv = train_test_split(X_train, y_train, test_size=0.33, stratify=y_train)
In [130]:
X_train, X_cv, y_train, y_cv = train_test_split(X_train, y_train, stratify= y_train,test_size = 0.33)
In [133]:
X_train.drop(["project_is_approved"], axis = 1, inplace = True)
X_test.drop(["project_is_approved"], axis = 1, inplace = True)
X_cv.drop(["project_is_approved"], axis = 1, inplace = True)
In [134]:
print(X_train.shape, y_train.shape)
print(X_cv.shape, y_cv.shape)
print(X_test.shape, y_test.shape)
print("="*100)
In [135]:
project_data.head(1)
Out[135]:
In [136]:
#Proprocessing for essay
# Combining all the above students
from tqdm import tqdm
preprocessed_essays_train = []
# tqdm is for printing the status bar
for sentance in tqdm(X_train['essay'].values):
sent = decontracted(sentance)
sent = sent.replace('\\r', ' ')
sent = sent.replace('\\"', ' ')
sent = sent.replace('\\n', ' ')
sent = re.sub('[^A-Za-z0-9]+', ' ', sent)
# https://gist.github.com/sebleier/554280
sent = ' '.join(e for e in sent.split() if e.lower() not in stopwords)
preprocessed_essays_train.append(sent.lower().strip())
In [137]:
#Proprocessing for essay
# Combining all the above students
from tqdm import tqdm
preprocessed_essays_test = []
# tqdm is for printing the status bar
for sentance in tqdm(X_test['essay'].values):
sent = decontracted(sentance)
sent = sent.replace('\\r', ' ')
sent = sent.replace('\\"', ' ')
sent = sent.replace('\\n', ' ')
sent = re.sub('[^A-Za-z0-9]+', ' ', sent)
# https://gist.github.com/sebleier/554280
sent = ' '.join(e for e in sent.split() if e.lower() not in stopwords)
preprocessed_essays_test.append(sent.lower().strip())
In [138]:
#Proprocessing for essay
# Combining all the above students
from tqdm import tqdm
preprocessed_essays_cv = []
# tqdm is for printing the status bar
for sentance in tqdm(X_cv['essay'].values):
sent = decontracted(sentance)
sent = sent.replace('\\r', ' ')
sent = sent.replace('\\"', ' ')
sent = sent.replace('\\n', ' ')
sent = re.sub('[^A-Za-z0-9]+', ' ', sent)
# https://gist.github.com/sebleier/554280
sent = ' '.join(e for e in sent.split() if e.lower() not in stopwords)
preprocessed_essays_cv.append(sent.lower().strip())
In [139]:
#Proprocessing for essay
# Combining all the above students
from tqdm import tqdm
preprocessed_titles_cv = []
# tqdm is for printing the status bar
for sentance in tqdm(X_cv['project_title'].values):
sent = decontracted(sentance)
sent = sent.replace('\\r', ' ')
sent = sent.replace('\\"', ' ')
sent = sent.replace('\\n', ' ')
sent = re.sub('[^A-Za-z0-9]+', ' ', sent)
# https://gist.github.com/sebleier/554280
sent = ' '.join(e for e in sent.split() if e.lower() not in stopwords)
preprocessed_titles_cv.append(sent.lower().strip())
In [140]:
#Proprocessing for essay
# Combining all the above students
from tqdm import tqdm
preprocessed_titles_train = []
# tqdm is for printing the status bar
for sentance in tqdm(X_train['project_title'].values):
sent = decontracted(sentance)
sent = sent.replace('\\r', ' ')
sent = sent.replace('\\"', ' ')
sent = sent.replace('\\n', ' ')
sent = re.sub('[^A-Za-z0-9]+', ' ', sent)
# https://gist.github.com/sebleier/554280
sent = ' '.join(e for e in sent.split() if e.lower() not in stopwords)
preprocessed_titles_train.append(sent.lower().strip())
In [141]:
#Proprocessing for essay
# Combining all the above students
from tqdm import tqdm
preprocessed_titles_test = []
# tqdm is for printing the status bar
for sentance in tqdm(X_test['project_title'].values):
sent = decontracted(sentance)
sent = sent.replace('\\r', ' ')
sent = sent.replace('\\"', ' ')
sent = sent.replace('\\n', ' ')
sent = re.sub('[^A-Za-z0-9]+', ' ', sent)
# https://gist.github.com/sebleier/554280
sent = ' '.join(e for e in sent.split() if e.lower() not in stopwords)
preprocessed_titles_test.append(sent.lower().strip())
1.5.1 Vectorizing Categorical data¶
In [142]:
# we use count vectorizer to convert the values into one
from sklearn.feature_extraction.text import CountVectorizer
vectorizer_clean = CountVectorizer(vocabulary=list(sorted_cat_dict.keys()), lowercase=False, binary=True)
vectorizer_clean.fit(X_train['clean_categories'].values)
X_train_cat = vectorizer_clean.transform(X_train['clean_categories'].values)
X_cv_cat = vectorizer_clean.transform(X_cv['clean_categories'].values)
X_test_cat = vectorizer_clean.transform(X_test['clean_categories'].values)
print(vectorizer_clean.get_feature_names())
print("After vectorizations")
print(X_train_cat.shape, y_train.shape)
print(X_cv_cat.shape, y_cv.shape)
print(X_test_cat.shape, y_test.shape)
feature_1 = vectorizer_clean.get_feature_names()
In [143]:
# we use count vectorizer to convert the values into one
vectorizer_clean_sub = CountVectorizer(vocabulary=list(sorted_sub_cat_dict.keys()), lowercase=False, binary=True)
vectorizer_clean_sub.fit(X_train['clean_subcategories'].values)
# we use the fitted CountVectorizer to convert the text to vector
X_train_clean_sub_ohe = vectorizer_clean_sub.transform(X_train['clean_subcategories'].values)
X_cv_clean_sub_ohe = vectorizer_clean_sub.transform(X_cv['clean_subcategories'].values)
X_test_clean_sub_ohe = vectorizer_clean_sub.transform(X_test['clean_subcategories'].values)
print(vectorizer_clean_sub.get_feature_names())
print("After vectorizations")
print(X_train_clean_sub_ohe.shape, y_train.shape)
print(X_cv_clean_sub_ohe.shape, y_cv.shape)
print(X_test_clean_sub_ohe.shape, y_test.shape)
feature_2 = vectorizer_clean_sub.get_feature_names()
In [144]:
# school_state convert categorical to vectors
# now time to cont the each words
from collections import Counter
my_counter = Counter()
for word in project_data['school_state'].values:
my_counter.update(word.split())# count the words
school_state_dict = dict(my_counter)# store in dicionary
sorted_school_state_dict = dict(sorted(school_state_dict.items(), key=lambda kv: kv[1]))
from sklearn.feature_extraction.text import CountVectorizer
vectorizer_state = CountVectorizer(vocabulary=list(sorted_school_state_dict.keys()), lowercase=False, binary=True)
vectorizer_state.fit(X_train['school_state'].values)
# firstly convert fit the train data into the vector then it learn the vocablery
# we use the fitted CountVectorizer to convert the text to vector
X_train_school_state = vectorizer_state.transform(X_train['school_state'].values)
X_cv_school_state = vectorizer_state.transform(X_cv['school_state'].values)
X_test_school_state = vectorizer_state.transform(X_test['school_state'].values)
print(vectorizer_state.get_feature_names())
print("After vectorizations")
print(X_train_school_state .shape, y_train.shape)
print(X_cv_school_state .shape, y_cv.shape)
print(X_test_school_state .shape, y_test.shape)
feature_3 = vectorizer_state.get_feature_names()
In [145]:
#https://stackoverflow.com/questions/42224700/attributeerror-float-object-has-no-attribute-split
project_data['clean_grade']=project_data['clean_grade'].fillna("")# fill the null values with space
# convert train,cv and test data of clean_categories into vectors
# we use count vectorizer to convert the values into one
from sklearn.feature_extraction.text import CountVectorizer
vectorizer_grade = CountVectorizer(vocabulary=list(sorted_project_grade_category_dict.keys()),lowercase=False, binary=True)
vectorizer_grade.fit(X_train['clean_grade'].values)
# firstly convert fit the train data into the vectoriaer then it learn hte vocablery
# we use the fitted CountVectorizer to convert the text to vector
X_train_project_grade_category = vectorizer_grade.transform(X_train['clean_grade'].values)
X_cv_project_grade_category = vectorizer_grade.transform(X_cv['clean_grade'].values)
X_test_project_grade_category = vectorizer_grade.transform(X_test['clean_grade'].values)
print(vectorizer_grade.get_feature_names())
print("After vectorizations")
print(X_train_project_grade_category .shape, y_train.shape)
print(X_cv_project_grade_category .shape, y_cv.shape)
print(X_test_project_grade_category .shape, y_test.shape)
feature_4 = vectorizer_grade.get_feature_names()
In [146]:
my_counter = Counter()
for teacher_prefix in project_data['teacher_prefix'].values:
teacher_prefix = str(teacher_prefix)
my_counter.update(teacher_prefix.split())
teacher_prefix_cat_dict = dict(my_counter)
sorted_teacher_prefix_cat_dict = dict(sorted(teacher_prefix_cat_dict.items(), key=lambda kv: kv[1]))
In [147]:
vectorizer_teacher = CountVectorizer(vocabulary=list(sorted_teacher_prefix_cat_dict.keys()), lowercase=False, binary=True)
vectorizer_teacher.fit(X_train['teacher_prefix'].values.astype("U"))
teacher_prefix_train = vectorizer_teacher.transform(X_train['teacher_prefix'].values.astype("U"))
teacher_prefix_test = vectorizer_teacher.transform(X_test['teacher_prefix'].values.astype("U"))
teacher_prefix_cv = vectorizer_teacher.transform(X_cv['teacher_prefix'].values.astype("U"))
In [148]:
print(vectorizer_teacher.get_feature_names())
print(teacher_prefix_train.shape,y_train.shape)
print(teacher_prefix_test.shape,y_test.shape)
print(teacher_prefix_cv.shape,y_cv.shape)
feature_5 = vectorizer_teacher.get_feature_names()
1.5.2 Vectorizing Text data¶
1.5.2.1 Bag of words¶
In [149]:
X_train_essay=preprocessed_essays_train
X_cv_essay=preprocessed_essays_cv
X_test_essay=preprocessed_essays_test
In [150]:
from sklearn.feature_extraction.text import CountVectorizer
vectorizer_b = CountVectorizer(min_df=10)
vectorizer_b.fit(X_train_essay)
Out[150]:
In [151]:
# we use the fitted CountVectorizer to convert the text to vector
X_train_essay_bow = vectorizer_b.transform(X_train_essay)
X_cv_essay_bow = vectorizer_b.transform(X_cv_essay)
X_test_essay_bow = vectorizer_b.transform(X_test_essay)
In [152]:
print(X_train_essay_bow.shape ,y_train.shape)
print(X_cv_essay_bow.shape ,y_train.shape)
print(X_test_essay_bow.shape ,y_train.shape)
In [153]:
X_train_title=preprocessed_titles_train
X_cv_title=preprocessed_titles_cv
X_test_title=preprocessed_titles_test
In [154]:
vectorizer_title_bow = CountVectorizer(min_df=10)
vectorizer_title_bow.fit(X_train_title)
Out[154]:
In [155]:
# we use the fitted CountVectorizer to convert the text to vector
X_train_bow_title = vectorizer_title_bow.transform(X_train_title)
X_cv_bow_title= vectorizer_title_bow.transform(X_cv_title)
X_test_bow_title = vectorizer_title_bow.transform(X_test_title)
In [156]:
print(X_train_bow_title.shape ,y_train.shape)
print(X_cv_bow_title.shape ,y_cv.shape)
print(X_test_bow_title.shape ,y_test.shape)
1.5.2.2 TFIDF vectorizer¶
In [157]:
from sklearn.feature_extraction.text import TfidfVectorizer
vectorizer_tfidf = TfidfVectorizer(min_df=10)
vectorizer_tfidf.fit(X_train_essay)
Out[157]:
In [158]:
# we use the fitted CountVectorizer to convert the text to vector
X_train_essay_tfidf = vectorizer_tfidf.transform(X_train_essay)
X_cv_essay_tfidf = vectorizer_tfidf.transform(X_cv_essay)
X_test_essay_tfidf = vectorizer_tfidf.transform(X_test_essay)
In [159]:
print(X_train_essay_tfidf.shape,y_train.shape)
print(X_cv_essay_tfidf.shape,y_cv.shape)
print(X_test_essay_tfidf.shape,y_test.shape)
In [160]:
vectorizer_tfidf_clean_titles = TfidfVectorizer(min_df=10)
vectorizer_tfidf_clean_titles.fit(X_train_title)
Out[160]:
In [161]:
# we use the fitted CountVectorizer to convert the text to vector
X_train_title_tfidf = vectorizer_tfidf_clean_titles.transform(X_train_title)
X_cv_title_tfidf = vectorizer_tfidf_clean_titles.transform(X_cv_title)
X_test_title_tfidf = vectorizer_tfidf_clean_titles.transform(X_test_title)
In [162]:
print(X_train_title_tfidf.shape,y_train.shape)
print(X_cv_title_tfidf.shape,y_cv.shape)
print(X_test_title_tfidf.shape,y_test.shape)
1.5.2.3 Using Pretrained Models: Avg W2V¶
In [163]:
'''
# Reading glove vectors in python: https://stackoverflow.com/a/38230349/4084039
def loadGloveModel(gloveFile):
print ("Loading Glove Model")
f = open(gloveFile,'r', encoding="utf8")
model = {}
for line in tqdm(f):
splitLine = line.split()
word = splitLine[0]
embedding = np.array([float(val) for val in splitLine[1:]])
model[word] = embedding
print ("Done.",len(model)," words loaded!")
return model
model = loadGloveModel('glove.42B.300d.txt')
# ============================
Output:
Loading Glove Model
1917495it [06:32, 4879.69it/s]
Done. 1917495 words loaded!
# ============================
words = []
for i in preproced_texts:
words.extend(i.split(' '))
for i in preproced_titles:
words.extend(i.split(' '))
print("all the words in the coupus", len(words))
words = set(words)
print("the unique words in the coupus", len(words))
inter_words = set(model.keys()).intersection(words)
print("The number of words that are present in both glove vectors and our coupus", \
len(inter_words),"(",np.round(len(inter_words)/len(words)*100,3),"%)")
words_courpus = {}
words_glove = set(model.keys())
for i in words:
if i in words_glove:
words_courpus[i] = model[i]
print("word 2 vec length", len(words_courpus))
# stronging variables into pickle files python: http://www.jessicayung.com/how-to-use-pickle-to-save-and-load-variables-in-python/
import pickle
with open('glove_vectors', 'wb') as f:
pickle.dump(words_courpus, f)
'''
Out[163]:
In [164]:
# stronging variables into pickle files python: http://www.jessicayung.com/how-to-use-pickle-to-save-and-load-variables-in-python/
# make sure you have the glove_vectors file
with open('glove_vectors (1)', 'rb') as f:
model = pickle.load(f)
glove_words = set(model.keys())
In [165]:
#for essay
# average Word2Vec
# compute average word2vec for each review.
def func(wordlist):
train_avg_w2v_vectors = []; # the avg-w2v for each sentence/review is stored in this list
for sentence in tqdm(wordlist): # for each review/sentence
vector = np.zeros(300) # as word vectors are of zero length # we are taking the 300dimensions very large
cnt_words =0; # num of words with a valid vector in the sentence/review
for word in sentence.split(): # for each word in a review/sentence
if word in glove_words:
vector += model[word]
cnt_words += 1
if cnt_words != 0:
vector /= cnt_words
train_avg_w2v_vectors.append(vector)
print(len(train_avg_w2v_vectors))
print(len(train_avg_w2v_vectors[0]))
return train_avg_w2v_vectors
In [166]:
train_avg_w2v_vectors=func(preprocessed_essays_train)
test_avg_w2v_vectors=func(preprocessed_essays_test)
cv_avg_w2v_vectors=func(preprocessed_essays_cv)
#for titles
train_avg_w2v_vectors_title=func(preprocessed_titles_train)
test_avg_w2v_vectors_title=func(preprocessed_titles_test)
cv_avg_w2v_vectors_title=func(preprocessed_titles_cv)
1.5.2.3 Using Pretrained Models: TFIDF weighted W2V¶
In [167]:
# S = ["abc def pqr", "def def def abc", "pqr pqr def"]
tfidf_model = TfidfVectorizer()
tfidf_model.fit(preprocessed_essays_train)
# we are converting a dictionary with word as a key, and the idf as a value
dictionary = dict(zip(tfidf_model.get_feature_names(), list(tfidf_model.idf_)))
tfidf_words = set(tfidf_model.get_feature_names())
In [168]:
# average Word2Vec
# compute average word2vec for each review.
def tf_idf(word_list):
train_title_tfidf_w2v_vectors = []; # the avg-w2v for each sentence/review is stored in this list
for sentence in tqdm(word_list): # for each review/sentence
vector = np.zeros(300) # as word vectors are of zero length
tf_idf_weight =0; # num of words with a valid vector in the sentence/review
for word in sentence.split():#.split(): # for each word in a review/sentence
if (word in glove_words) and (word in tfidf_words):
#vec = model.wv[word]
vec = model[word] # getting the vector for each word
# here we are multiplying idf value(dictionary[word]) and the tf value((sentence.count(word)/len(sentence.split())))
tf_idf = dictionary[word]*(sentence.count(word)/len(sentence.split()))
vector += (vec * tf_idf) # calculating tfidf weighted w2v
tf_idf_weight += tf_idf
if tf_idf_weight != 0:
vector /= tf_idf_weight
train_title_tfidf_w2v_vectors.append(vector)
print(len(train_title_tfidf_w2v_vectors))
print(len(train_title_tfidf_w2v_vectors[0]))
return train_title_tfidf_w2v_vectors
In [169]:
#train_title_tfidf_w2v_vector
train_tfidf_w2v_vectors=tf_idf(preprocessed_essays_train)
test_tfidf_w2v_vectors=tf_idf(preprocessed_essays_test)
cv_tfidf_w2v_vectors=tf_idf(preprocessed_essays_cv)
#train_title_tfidf_w2v_vector
train_title_tfidf_w2v_vectors=tf_idf(preprocessed_titles_train)
test_title_tfidf_w2v_vectors=tf_idf(preprocessed_titles_test)
cv_title_tfidf_w2v_vectors=tf_idf(preprocessed_titles_cv)
1.5.3 Vectorizing Numerical features¶
In [170]:
# https://stackoverflow.com/questions/22407798/how-to-reset-a-dataframes-indexes-for-all-groups-in-one-step
price_data = resource_data.groupby('id').agg({'price':'sum', 'quantity':'sum'}).reset_index()
df = pd.merge(project_data, price_data, on='id', how='left')
print(price_data.head(2))
In [171]:
# join two dataframes in python:
X_train = pd.merge(X_train, price_data, on='id', how='left')
X_test = pd.merge(X_test, price_data, on='id', how='left')
X_cv = pd.merge(X_cv, price_data, on='id', how='left')
In [172]:
#standardization
# check this one: https://www.youtube.com/watch?v=0HOqOcln3Z4&t=530s
# standardization sklearn: https://scikitlearn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import MinMaxScaler
from sklearn import preprocessing
price_scalar = StandardScaler()
price_scalar.fit(X_train['price'].values.reshape(-1,1)) # finding the mean and standard deviationof this data
#print(f"Mean : {price_scalar.mean_[0]}, Standard deviation : {np.sqrt(price_scalar.var_[0])}")
# Now standardize the data with above mean and variance.
train_price = price_scalar.transform(X_train['price'].values.reshape(-1, 1))
test_price = price_scalar.transform(X_test['price'].values.reshape(-1, 1))
cv_price = price_scalar.transform(X_cv['price'].values.reshape(-1, 1))
print(train_price.shape,y_train.shape)
print(test_price.shape,y_train.shape)
print(cv_price.shape,y_cv.shape)
In [173]:
# previous_year_projects
price_scalar.fit(X_train['teacher_number_of_previously_posted_projects'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
#print(f"Mean : {price_scalar.mean_[0]}, Standard deviation : {np.sqrt(price_scalar.var_[0])}")
# Now standardize the data with above maen and variance.
train_T =price_scalar.transform(X_train['teacher_number_of_previously_posted_projects'].values.reshape(-1,1))
test_T =price_scalar.transform(X_test['teacher_number_of_previously_posted_projects'].values.reshape(-1, 1))
cv_T = price_scalar.transform(X_cv['teacher_number_of_previously_posted_projects'].values.reshape(-1, 1))
In [174]:
print(train_T.shape,y_train.shape)
print(test_T.shape,y_train.shape)
print(cv_T.shape,y_cv.shape)
In [175]:
price_scalar.fit(X_train['quantity'].values.reshape(-1,1)) # finding the mean and standarddeviation of this data
#print(f"Mean : {price_scalar.mean_[0]}, Standard deviation : {np.sqrt(price_scalar.var_[0])}")
# Now standardize the data with above maen and variance.
train_quantity = price_scalar.transform(X_train['quantity'].values.reshape(-1, 1))
cv_quantity = price_scalar.transform(X_cv['quantity'].values.reshape(-1, 1))
test_quantity = price_scalar.transform(X_test['quantity'].values.reshape(-1, 1))
print(train_quantity .shape,y_train.shape)
print(test_quantity .shape,y_train.shape)
print(cv_quantity .shape,y_cv.shape)
1.5.4 Merging all the above features¶
- we need to merge all the numerical vectors i.e catogorical, text, numerical vectors
categorical, numerical features + project_title(BOW) + preprocessed_eassay (BOW)¶
In [176]:
from scipy.sparse import hstack
X_tr = hstack((X_train_essay_bow,X_train_bow_title,X_train_cat,X_train_clean_sub_ohe,X_train_school_state,X_train_project_grade_category,teacher_prefix_train,train_price,train_T,train_quantity)).tocsr()
X_te = hstack((X_test_essay_bow,X_test_bow_title,X_test_cat,X_test_clean_sub_ohe,X_test_school_state,X_test_project_grade_category,teacher_prefix_test,test_price,test_T,test_quantity)).tocsr()
X_cr = hstack((X_cv_essay_bow,X_cv_bow_title,X_cv_cat,X_cv_clean_sub_ohe,X_cv_school_state,X_cv_project_grade_category,teacher_prefix_cv,cv_price,cv_T,cv_quantity )).tocsr()
In [177]:
print(X_tr.shape,y_train.shape)
print(X_te.shape,y_test.shape)
print(X_cr.shape,y_cv.shape)
In [178]:
#Feature aggregation
f=vectorizer_b.get_feature_names()
g=vectorizer_title_bow.get_feature_names()
h=vectorizer_tfidf.get_feature_names()
i=vectorizer_tfidf_clean_titles.get_feature_names()
feature_agg_bow = feature_1 + feature_2 + feature_3 + feature_4 + feature_5 + f + g
feature_agg_tfidf = feature_1 + feature_2 + feature_3+ feature_4 + feature_5 +h + i
# p is price, q is quantity, t is teacher previous year projects
feature_agg_bow.append('price')
feature_agg_tfidf.append('price')
feature_agg_bow.append('quantity')
feature_agg_tfidf.append('quantity')
feature_agg_bow.append('teacher_previous_projects')
feature_agg_tfidf.append('teacher_previous_projects')
In [179]:
len(feature_agg_bow)
Out[179]:
In [180]:
from sklearn.tree import DecisionTreeClassifier
dtree = DecisionTreeClassifier(max_depth = 3)
clf = dtree.fit(X_tr, y_train)
In [181]:
import os
os.environ["PATH"] += os.pathsep + 'C:/Program Files (x86)/Graphviz2.38/bin/'
In [182]:
import warnings
warnings.filterwarnings("ignore")
from sklearn.externals.six import StringIO
from IPython.display import Image
from sklearn.tree import export_graphviz
import pydotplus
dot_data = StringIO()
export_graphviz(clf, out_file=dot_data, filled=True, rounded=True, special_characters=True, feature_names=feature_agg_bow,rotate=True)
graph = pydotplus.graph_from_dot_data(dot_data.getvalue())
Image(graph.create_png())
Out[182]:
In [183]:
from sklearn.model_selection import GridSearchCV
from sklearn.tree import DecisionTreeClassifier
dtree_1 = DecisionTreeClassifier(class_weight = 'balanced')
parameters = {'max_depth':[1, 5, 10, 50, 100, 500, 100],'min_samples_split':[5, 10, 100, 500]}
In [184]:
clf_1 = GridSearchCV(dtree_1,parameters,cv =3,scoring = 'roc_auc',return_train_score = True)
clf_1 = clf_1.fit(X_tr,y_train)
In [185]:
train_auc= clf_1.cv_results_['mean_train_score']
train_auc_std= clf_1.cv_results_['std_train_score']
cv_auc = clf_1.cv_results_['mean_test_score']
cv_auc_std= clf_1.cv_results_['std_test_score']
In [186]:
train_auc
Out[186]:
In [187]:
cv_auc
Out[187]:
In [188]:
print(clf_1.best_estimator_)
#Mean cross-validated score of the best_estimator
print(clf_1.score(X_tr,y_train))
print(clf_1.score(X_te,y_test))
In [189]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [190]:
# https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_curve.html#sklearn.metrics.roc_curve
from sklearn.metrics import roc_curve, auc
model = DecisionTreeClassifier(max_depth = 10, min_samples_split = 500)
model.fit(X_tr, y_train)
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_train_pred = batch_predict(model, X_tr)
y_test_pred = batch_predict(model, X_te)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(TPR)")
plt.ylabel("True Positive Rate(FPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [191]:
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(fpr*(1-tpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
global prediction
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
prediction= predictions
return predictions
In [192]:
print("="*100)
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train, predict(y_train_pred, tr_thresholds, train_fpr, train_fpr)))
In [193]:
conf_matr_df_train_1 = pd.DataFrame(confusion_matrix(y_train, predict(y_train_pred, tr_thresholds, train_fpr, train_fpr)), range(2),range(2))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_train_1, annot=True,annot_kws={"size": 16}, fmt='g')
Out[193]:
In [194]:
print("Test confusion matrix")
print(confusion_matrix(y_test, predict(y_test_pred, tr_thresholds, test_fpr, test_fpr)))
In [195]:
conf_matr_df_test_1 = pd.DataFrame(confusion_matrix(y_test, predict(y_test_pred, tr_thresholds, test_fpr, test_fpr)), range(2),range(2))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_test_1, annot=True,annot_kws={"size": 16}, fmt='g')
Out[195]:
In [199]:
fpi = []
for i in range(len(y_test)):
if (y_test.values[i] == 0 ) & (prediction[i] == 1):
fpi.append(i)
fp_essay1 = []
for i in fpi:
fp_essay1.append(X_test['essay'].values[i])
In [200]:
from wordcloud import WordCloud, STOPWORDS
comment_words = ' '
stopwords = set(STOPWORDS)
for val in fp_essay1 :
val = str(val)
tokens = val.split()
for i in range(len(tokens)):
tokens[i] = tokens[i].lower()
for words in tokens :
comment_words = comment_words + words + ' '
wordcloud = WordCloud(width = 800, height = 800, background_color ='white', stopwords = stopwords,min_font_size = 10).generate(comment_words)
plt.figure(figsize = (6, 6), facecolor = None)
plt.imshow(wordcloud)
plt.axis("off")
plt.tight_layout(pad = 0)
plt.show()
In [203]:
# first get the columns:
cols = X_test.columns
X_test_falsePos1 = pd.DataFrame(columns=cols)
# get the data of the false pisitives
for i in fpi : # (in fpi all the false positives data points indexes)
X_test_falsePos1 = X_test_falsePos1.append(X_test.filter(items=[i], axis=0))
X_test_falsePos1.head(1)
len(X_test_falsePos1)
Out[203]:
In [204]:
##Box Plot (FP 'price')
sns.boxplot(y='price', data=X_test_falsePos1)
Out[204]:
In [205]:
##PDF (FP ,teacher_number_of_previously_posted_projects)
plt.figure(figsize=(8,5))
counts, bin_edges = np.histogram(X_test_falsePos1['teacher_number_of_previously_posted_projects'],
bins='auto', density=True)
pdf = counts/sum(counts)
pdfP, = plt.plot(bin_edges[1:], pdf)
plt.legend([pdfP], ["PDF"])
plt.xlabel('teacher_number_of_previously_posted_projects')
plt.show
Out[205]:
In [215]:
c=DecisionTreeClassifier (class_weight = 'balanced',criterion="entropy",max_depth=10,min_samples_split=500)
c.fit(X_tr, y_train)
Out[215]:
In [ ]:
#Graphviz visualization of Decision Tree on bow
In [216]:
import warnings
warnings.filterwarnings("ignore")
from sklearn.externals.six import StringIO
from IPython.display import Image
from sklearn.tree import export_graphviz
import pydotplus
dot_data = StringIO()
export_graphviz(c, out_file=dot_data, filled=True, rounded=True, special_characters=True, feature_names=feature_agg_bow,rotate=True)
graph = pydotplus.graph_from_dot_data(dot_data.getvalue())
Image(graph.create_png())
Out[216]:
categorical, numerical features + project_title(tfidf) + preprocessed_eassay (tfidf)¶
In [217]:
from scipy.sparse import hstack
X_tr_tfidf = hstack((X_train_essay_tfidf,X_train_title_tfidf,X_train_cat,X_train_clean_sub_ohe,X_train_school_state,X_train_project_grade_category,teacher_prefix_train,train_price,train_T,train_quantity)).tocsr()
X_te_tfidf = hstack((X_test_essay_tfidf,X_test_title_tfidf,X_test_cat,X_test_clean_sub_ohe,X_test_school_state,X_test_project_grade_category,teacher_prefix_test,test_price,test_T,test_quantity)).tocsr()
X_cr_tfidf = hstack((X_cv_essay_tfidf,X_cv_title_tfidf,X_cv_cat,X_cv_clean_sub_ohe,X_cv_school_state,X_cv_project_grade_category,teacher_prefix_cv,cv_price,cv_T,cv_quantity )).tocsr()
In [218]:
print(X_tr_tfidf.shape,y_train.shape)
print(X_te_tfidf.shape,y_test.shape)
print(X_cr_tfidf.shape,y_cv.shape)
In [220]:
#Feature aggregation
f=vectorizer_b.get_feature_names()
g=vectorizer_title_bow.get_feature_names()
h=vectorizer_tfidf.get_feature_names()
i=vectorizer_tfidf_clean_titles.get_feature_names()
feature_agg_bow = feature_1 + feature_2 + feature_3 + feature_4 + feature_5 + f + g
feature_agg_tfidf = feature_1 + feature_2 + feature_3+ feature_4 + feature_5 +h + i
# p is price, q is quantity, t is teacher previous year projects
feature_agg_bow.append('price')
feature_agg_tfidf.append('price')
feature_agg_bow.append('quantity')
feature_agg_tfidf.append('quantity')
feature_agg_bow.append('teacher_previous_projects')
feature_agg_tfidf.append('teacher_previous_projects')
In [221]:
len(feature_agg_tfidf)
Out[221]:
In [222]:
from sklearn.tree import DecisionTreeClassifier
dtree_1 = DecisionTreeClassifier(max_depth = 3)
clf_1 = dtree_1.fit(X_tr_tfidf, y_train)
In [223]:
import os
os.environ["PATH"] += os.pathsep + 'C:/Program Files (x86)/Graphviz2.38/bin/'
In [224]:
import warnings
warnings.filterwarnings("ignore")
from sklearn.externals.six import StringIO
from IPython.display import Image
from sklearn.tree import export_graphviz
import pydotplus
dot_data = StringIO()
export_graphviz(clf_1, out_file=dot_data, filled=True, rounded=True, special_characters=True, feature_names=feature_agg_tfidf,rotate=True)
graph = pydotplus.graph_from_dot_data(dot_data.getvalue())
Image(graph.create_png())
Out[224]:
In [225]:
from sklearn.model_selection import GridSearchCV
from sklearn.tree import DecisionTreeClassifier
dtree_1 = DecisionTreeClassifier(class_weight = 'balanced')
parameters = {'max_depth':[1, 5, 10, 50, 100, 500, 100],'min_samples_split':[5, 10, 100, 500]}
In [226]:
clf_1 = GridSearchCV(dtree_1,parameters,cv =3,scoring = 'roc_auc',return_train_score = True)
clf_1 = clf_1.fit(X_tr_tfidf,y_train)
In [227]:
train_auc= clf_1.cv_results_['mean_train_score']
train_auc_std= clf_1.cv_results_['std_train_score']
cv_auc = clf_1.cv_results_['mean_test_score']
cv_auc_std= clf_1.cv_results_['std_test_score']
In [228]:
train_auc
Out[228]:
In [229]:
cv_auc
Out[229]:
In [230]:
print(clf_1.best_estimator_)
#Mean cross-validated score of the best_estimator
print(clf_1.score(X_tr_tfidf,y_train))
print(clf_1.score(X_te_tfidf,y_test))
In [231]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [232]:
# https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_curve.html#sklearn.metrics.roc_curve
from sklearn.metrics import roc_curve, auc
model = DecisionTreeClassifier(max_depth = 10, min_samples_split = 500)
model.fit(X_tr_tfidf, y_train)
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_train_pred = batch_predict(model, X_tr_tfidf)
y_test_pred = batch_predict(model, X_te_tfidf)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(TPR)")
plt.ylabel("True Positive Rate(FPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [233]:
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(fpr*(1-tpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
global prediction
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
prediction= predictions
return predictions
In [234]:
print("="*100)
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train, predict(y_train_pred, tr_thresholds, train_fpr, train_fpr)))
In [235]:
conf_matr_df_train_1 = pd.DataFrame(confusion_matrix(y_train, predict(y_train_pred, tr_thresholds, train_fpr, train_fpr)), range(2),range(2))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_train_1, annot=True,annot_kws={"size": 16}, fmt='g')
Out[235]:
In [236]:
print("Test confusion matrix")
print(confusion_matrix(y_test, predict(y_test_pred, tr_thresholds, test_fpr, test_fpr)))
In [237]:
conf_matr_df_test_1 = pd.DataFrame(confusion_matrix(y_test, predict(y_test_pred, tr_thresholds, test_fpr, test_fpr)), range(2),range(2))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_test_1, annot=True,annot_kws={"size": 16}, fmt='g')
Out[237]:
In [245]:
fpi = []
for i in range(len(y_test)):
if (y_test.values[i] == 0 ) & (prediction[i] == 1):
fpi.append(i)
fp_essay1 = []
for i in fpi:
fp_essay1.append(X_test['essay'].values[i])
In [246]:
from wordcloud import WordCloud, STOPWORDS
comment_words = ' '
stopwords = set(STOPWORDS)
for val in fp_essay1 :
val = str(val)
tokens = val.split()
for i in range(len(tokens)):
tokens[i] = tokens[i].lower()
for words in tokens :
comment_words = comment_words + words + ' '
wordcloud = WordCloud(width = 800, height = 800, background_color ='white', stopwords = stopwords,min_font_size = 10).generate(comment_words)
plt.figure(figsize = (6, 6), facecolor = None)
plt.imshow(wordcloud)
plt.axis("off")
plt.tight_layout(pad = 0)
plt.show()
In [247]:
# first get the columns:
cols = X_test.columns
X_test_falsePos1 = pd.DataFrame(columns=cols)
# get the data of the false pisitives
for i in fpi : # (in fpi all the false positives data points indexes)
X_test_falsePos1 = X_test_falsePos1.append(X_test.filter(items=[i], axis=0))
X_test_falsePos1.head(1)
len(X_test_falsePos1)
Out[247]:
In [248]:
##Box Plot (FP 'price')
sns.boxplot(y='price', data=X_test_falsePos1)
Out[248]:
In [249]:
##PDF (FP ,teacher_number_of_previously_posted_projects)
plt.figure(figsize=(8,5))
counts, bin_edges = np.histogram(X_test_falsePos1['teacher_number_of_previously_posted_projects'],
bins='auto', density=True)
pdf = counts/sum(counts)
pdfP, = plt.plot(bin_edges[1:], pdf)
plt.legend([pdfP], ["PDF"])
plt.xlabel('teacher_number_of_previously_posted_projects')
plt.show
Out[249]:
Graphviz visualization of Decision Tree on TFIDF,¶
In [250]:
import warnings
warnings.filterwarnings("ignore")
from sklearn.externals.six import StringIO
from IPython.display import Image
from sklearn.tree import export_graphviz
import pydotplus
dot_data = StringIO()
export_graphviz(model, out_file=dot_data, filled=True, rounded=True, special_characters=True, feature_names=feature_agg_tfidf,rotate=True)
graph = pydotplus.graph_from_dot_data(dot_data.getvalue())
Image(graph.create_png())
Out[250]:
categorical, numerical features + project_title(avgw2v) + preprocessed_eassay (avgw2v)¶
In [251]:
from scipy.sparse import hstack
X_tr_w2v = hstack((train_avg_w2v_vectors,train_avg_w2v_vectors_title ,X_train_cat,X_train_clean_sub_ohe,X_train_school_state,X_train_project_grade_category,teacher_prefix_train,train_price,train_T,train_quantity)).tocsr()
X_te_w2v = hstack((test_avg_w2v_vectors,test_avg_w2v_vectors_title,X_test_cat,X_test_clean_sub_ohe,X_test_school_state,X_test_project_grade_category,teacher_prefix_test,test_price,test_T,test_quantity)).tocsr()
X_cr_w2v = hstack((cv_avg_w2v_vectors,cv_avg_w2v_vectors_title,X_cv_cat,X_cv_clean_sub_ohe,X_cv_school_state,X_cv_project_grade_category,teacher_prefix_cv,cv_price,cv_T,cv_quantity )).tocsr()
In [252]:
print(X_tr_w2v.shape,y_train.shape)
print(X_te_w2v.shape,y_test.shape)
print(X_cr_w2v.shape,y_cv.shape)
In [253]:
from sklearn.model_selection import GridSearchCV
from sklearn.tree import DecisionTreeClassifier
dtree_1 = DecisionTreeClassifier(class_weight = 'balanced')
parameters = {'max_depth':[1, 5, 10, 50, 100, 500, 100],'min_samples_split':[5, 10,25,50 ,100, 500]}
In [254]:
clf_1 = GridSearchCV(dtree_1,parameters,cv =3,scoring = 'roc_auc',return_train_score = True)
clf_1 = clf_1.fit(X_tr_w2v,y_train)
In [255]:
train_auc= clf_1.cv_results_['mean_train_score']
train_auc_std= clf_1.cv_results_['std_train_score']
cv_auc = clf_1.cv_results_['mean_test_score']
cv_auc_std= clf_1.cv_results_['std_test_score']
In [256]:
train_auc
Out[256]:
In [257]:
cv_auc
Out[257]:
In [258]:
print(clf_1.best_estimator_)
#Mean cross-validated score of the best_estimator
print(clf_1.score(X_tr_w2v,y_train))
print(clf_1.score(X_te_w2v,y_test))
In [259]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [291]:
# https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_curve.html#sklearn.metrics.roc_curve
from sklearn.metrics import roc_curve, auc
model = DecisionTreeClassifier(max_depth = 5, min_samples_split = 500)
model.fit(X_tr_w2v, y_train)
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_train_pred = batch_predict(model, X_tr_w2v)
y_test_pred = batch_predict(model, X_te_w2v)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(TPR)")
plt.ylabel("True Positive Rate(FPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [292]:
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(fpr*(1-tpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
global prediction
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
prediction= predictions
return predictions
In [293]:
print("="*100)
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train, predict(y_train_pred, tr_thresholds, train_fpr, train_fpr)))
In [294]:
conf_matr_df_train_1 = pd.DataFrame(confusion_matrix(y_train, predict(y_train_pred, tr_thresholds, train_fpr, train_fpr)), range(2),range(2))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_train_1, annot=True,annot_kws={"size": 16}, fmt='g')
Out[294]:
In [295]:
print("Test confusion matrix")
print(confusion_matrix(y_test, predict(y_test_pred, tr_thresholds, test_fpr, test_fpr)))
In [296]:
conf_matr_df_test_1 = pd.DataFrame(confusion_matrix(y_test, predict(y_test_pred, tr_thresholds, test_fpr, test_fpr)), range(2),range(2))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_test_1, annot=True,annot_kws={"size": 16}, fmt='g')
Out[296]:
In [297]:
fpi = []
for i in range(len(y_test)):
if (y_test.values[i] == 0 ) & (prediction[i] == 1):
fpi.append(i)
fp_essay1 = []
for i in fpi:
fp_essay1.append(X_test['essay'].values[i])
In [298]:
from wordcloud import WordCloud, STOPWORDS
comment_words = ' '
stopwords = set(STOPWORDS)
for val in fp_essay1 :
val = str(val)
tokens = val.split()
for i in range(len(tokens)):
tokens[i] = tokens[i].lower()
for words in tokens :
comment_words = comment_words + words + ' '
wordcloud = WordCloud(width = 800, height = 800, background_color ='white', stopwords = stopwords,min_font_size = 10).generate(comment_words)
plt.figure(figsize = (6, 6), facecolor = None)
plt.imshow(wordcloud)
plt.axis("off")
plt.tight_layout(pad = 0)
plt.show()
In [299]:
# first get the columns:
cols = X_test.columns
X_test_falsePos1 = pd.DataFrame(columns=cols)
# get the data of the false pisitives
for i in fpi : # (in fpi all the false positives data points indexes)
X_test_falsePos1 = X_test_falsePos1.append(X_test.filter(items=[i], axis=0))
X_test_falsePos1.head(1)
len(X_test_falsePos1)
Out[299]:
In [300]:
##Box Plot (FP 'price')
sns.boxplot(y='price', data=X_test_falsePos1)
Out[300]:
In [301]:
##PDF (FP ,teacher_number_of_previously_posted_projects)
plt.figure(figsize=(8,5))
counts, bin_edges = np.histogram(X_test_falsePos1['teacher_number_of_previously_posted_projects'],
bins='auto', density=True)
pdf = counts/sum(counts)
pdfP, = plt.plot(bin_edges[1:], pdf)
plt.legend([pdfP], ["PDF"])
plt.xlabel('teacher_number_of_previously_posted_projects')
plt.show
Out[301]:
categorical, numerical features + project_title(tfidfw2v) + preprocessed_eassay (tfidfw2v)¶
In [281]:
from scipy.sparse import hstack
X_tr_tfidf_w2v = hstack((train_tfidf_w2v_vectors,train_title_tfidf_w2v_vectors,X_train_cat,X_train_clean_sub_ohe,X_train_school_state,X_train_project_grade_category,teacher_prefix_train,train_price,train_T,train_quantity)).tocsr()
X_te_tfidf_w2v = hstack((test_tfidf_w2v_vectors,test_title_tfidf_w2v_vectors,X_test_cat,X_test_clean_sub_ohe,X_test_school_state,X_test_project_grade_category,teacher_prefix_test,test_price,test_T,test_quantity)).tocsr()
X_cr_tfidf_w2v = hstack((cv_tfidf_w2v_vectors,cv_title_tfidf_w2v_vectors,X_cv_cat,X_cv_clean_sub_ohe,X_cv_school_state,X_cv_project_grade_category,teacher_prefix_cv,cv_price,cv_T,cv_quantity )).tocsr()
In [282]:
print(X_tr_tfidf_w2v.shape,y_train.shape)
print(X_te_tfidf_w2v.shape,y_test.shape)
print(X_cr_tfidf_w2v.shape,y_cv.shape)
In [283]:
from sklearn.model_selection import GridSearchCV
from sklearn.tree import DecisionTreeClassifier
dtree_1 = DecisionTreeClassifier()
parameters = {'max_depth':[1, 5, 10, 50, 100, 500, 100],'min_samples_split':[5, 10,25,50,75, 100, 500]}
In [284]:
clf_1 = GridSearchCV(dtree_1,parameters,cv =3,scoring = 'roc_auc',return_train_score = True)
clf_1 = clf_1.fit(X_tr_tfidf_w2v,y_train)
In [285]:
train_auc= clf_1.cv_results_['mean_train_score']
train_auc_std= clf_1.cv_results_['std_train_score']
cv_auc = clf_1.cv_results_['mean_test_score']
cv_auc_std= clf_1.cv_results_['std_test_score']
In [286]:
train_auc
Out[286]:
In [287]:
cv_auc
Out[287]:
In [288]:
print(clf_1.best_estimator_)
#Mean cross-validated score of the best_estimator
print(clf_1.score(X_tr_tfidf_w2v,y_train))
print(clf_1.score(X_te_tfidf_w2v,y_test))
In [289]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [302]:
# https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_curve.html#sklearn.metrics.roc_curve
from sklearn.metrics import roc_curve, auc
model = DecisionTreeClassifier(max_depth = 5, min_samples_split = 500)
model.fit(X_tr_tfidf_w2v, y_train)
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_train_pred = batch_predict(model, X_tr_tfidf_w2v)
y_test_pred = batch_predict(model, X_te_tfidf_w2v)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(TPR)")
plt.ylabel("True Positive Rate(FPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [303]:
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(fpr*(1-tpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
global prediction
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
prediction= predictions
return predictions
In [304]:
print("="*100)
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train, predict(y_train_pred, tr_thresholds, train_fpr, train_fpr)))
In [305]:
conf_matr_df_train_1 = pd.DataFrame(confusion_matrix(y_train, predict(y_train_pred, tr_thresholds, train_fpr, train_fpr)), range(2),range(2))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_train_1, annot=True,annot_kws={"size": 16}, fmt='g')
Out[305]:
In [306]:
print("Test confusion matrix")
print(confusion_matrix(y_test, predict(y_test_pred, tr_thresholds, test_fpr, test_fpr)))
In [307]:
conf_matr_df_test_1 = pd.DataFrame(confusion_matrix(y_test, predict(y_test_pred, tr_thresholds, test_fpr, test_fpr)), range(2),range(2))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_test_1, annot=True,annot_kws={"size": 16}, fmt='g')
Out[307]:
In [308]:
fpi = []
for i in range(len(y_test)):
if (y_test.values[i] == 0 ) & (prediction[i] == 1):
fpi.append(i)
fp_essay1 = []
for i in fpi:
fp_essay1.append(X_test['essay'].values[i])
In [309]:
from wordcloud import WordCloud, STOPWORDS
comment_words = ' '
stopwords = set(STOPWORDS)
for val in fp_essay1 :
val = str(val)
tokens = val.split()
for i in range(len(tokens)):
tokens[i] = tokens[i].lower()
for words in tokens :
comment_words = comment_words + words + ' '
wordcloud = WordCloud(width = 800, height = 800, background_color ='white', stopwords = stopwords,min_font_size = 10).generate(comment_words)
plt.figure(figsize = (6, 6), facecolor = None)
plt.imshow(wordcloud)
plt.axis("off")
plt.tight_layout(pad = 0)
plt.show()
In [310]:
# first get the columns:
cols = X_test.columns
X_test_falsePos1 = pd.DataFrame(columns=cols)
# get the data of the false pisitives
for i in fpi : # (in fpi all the false positives data points indexes)
X_test_falsePos1 = X_test_falsePos1.append(X_test.filter(items=[i], axis=0))
X_test_falsePos1.head(1)
len(X_test_falsePos1)
Out[310]:
In [311]:
##Box Plot (FP 'price')
sns.boxplot(y='price', data=X_test_falsePos1)
Out[311]:
In [312]:
##PDF (FP ,teacher_number_of_previously_posted_projects)
plt.figure(figsize=(8,5))
counts, bin_edges = np.histogram(X_test_falsePos1['teacher_number_of_previously_posted_projects'],
bins='auto', density=True)
pdf = counts/sum(counts)
pdfP, = plt.plot(bin_edges[1:], pdf)
plt.legend([pdfP], ["PDF"])
plt.xlabel('teacher_number_of_previously_posted_projects')
plt.show
Out[312]:
3. Conclusion
In [313]:
# http://zetcode.com/python/prettytable/
from prettytable import PrettyTable
#If you get a ModuleNotFoundError error , install prettytable using: pip3 install prettytable
x = PrettyTable()
x.field_names = ["Vectorizer", "Model", "Hyperparameters(max depth,min samples split)", "Train AUC", "Test AUC"]
x.add_row(["BOW", "Decision Trees","(10, 100)", 0.711, 0.664])
x.add_row(["TFIDF", "Decision Trees", "(10, 500)", 0.729, 0.658])
x.add_row(["AVG W2V", "Decision Trees", "(5, 500)", 0.671, 0.650])
x.add_row(["TFIDF W2V", "Decision Trees", "(5, 50)", 0.680, 0.643])
print(x)